@umituz/react-native-ai-generation-content 1.84.6 → 1.84.8

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@umituz/react-native-ai-generation-content",
-  "version": "1.84.6",
+  "version": "1.84.8",
   "description": "Provider-agnostic AI generation orchestration for React Native with result preview components",
   "main": "src/index.ts",
   "types": "src/index.ts",

package/src/domain/interfaces/video-model-config.types.ts

@@ -37,6 +37,7 @@ export interface VideoModelConfig {
     readonly duration?: number;
     readonly aspectRatio?: string;
     readonly resolution?: string;
+    readonly audioUrl?: string;
   }) => Record<string, unknown>;
 
   /** Pricing data for credit calculation (keys = resolution IDs) */

package/src/domains/generation/wizard/infrastructure/strategies/video-generation.executor.ts

@@ -50,6 +50,9 @@ function buildGenericInput(input: WizardVideoInput): Record<string, unknown> {
   if (input.resolution) {
     modelInput.resolution = input.resolution;
   }
+  if (input.audioUrl) {
+    modelInput.audio = input.audioUrl;
+  }
 
   return modelInput;
 }

package/src/domains/generation/wizard/infrastructure/strategies/video-generation.strategy.ts

@@ -12,9 +12,32 @@ import { extractPhotosAsBase64 } from "./shared/photo-extraction.utils";
 import type { WizardVideoInput, CreateVideoStrategyOptions } from "./video-generation.types";
 import { validatePhotoCount, validateWizardVideoInput } from "./video-generation.types";
 import { executeVideoGeneration, submitVideoGenerationToQueue } from "./video-generation.executor";
+import { readAsStringAsync, EncodingType } from "expo-file-system";
 
 
 
+/**
+ * Extract audio from wizardData and read as base64.
+ * Audio step stores data as { uri: "file:///..." }.
+ */
+async function extractAudioAsBase64(wizardData: Record<string, unknown>): Promise<string | undefined> {
+  const audioData = wizardData.background_audio as { uri?: string } | undefined;
+  if (!audioData?.uri) return undefined;
+
+  try {
+    const base64 = await readAsStringAsync(audioData.uri, { encoding: EncodingType.Base64 });
+    if (typeof __DEV__ !== "undefined" && __DEV__) {
+      console.log("[VideoStrategy] Audio extracted as base64", { length: base64.length });
+    }
+    return base64;
+  } catch (error) {
+    if (typeof __DEV__ !== "undefined" && __DEV__) {
+      console.warn("[VideoStrategy] Failed to read audio file:", error);
+    }
+    return undefined;
+  }
+}
+
 export async function buildVideoInput(
   wizardData: Record<string, unknown>,
   scenario: WizardScenarioData,
@@ -23,6 +46,29 @@ export async function buildVideoInput(
     console.log("[VideoStrategy] Building input", { scenarioId: scenario.id });
   }
 
+  // Extract audio (shared by all code paths)
+  const audioUrl = await extractAudioAsBase64(wizardData);
+
+  // If a pre-generated image URL exists (e.g., two-step solo video),
+  // use it directly instead of extracting photos from wizard data
+  if (scenario.preGeneratedImageUrl) {
+    if (typeof __DEV__ !== "undefined" && __DEV__) {
+      console.log("[VideoStrategy] Using pre-generated image URL", {
+        url: scenario.preGeneratedImageUrl.slice(0, 80),
+        hasAudio: !!audioUrl,
+      });
+    }
+    const finalPrompt = extractPrompt(wizardData, scenario.aiPrompt) || scenario.aiPrompt || "";
+    return {
+      sourceImageBase64: scenario.preGeneratedImageUrl,
+      prompt: finalPrompt,
+      duration: extractDuration(wizardData),
+      aspectRatio: extractAspectRatio(wizardData),
+      resolution: extractResolution(wizardData),
+      audioUrl,
+    };
+  }
+
   const photos = await extractPhotosAsBase64(wizardData, true);
 
   const validation = validatePhotoCount(photos.length, scenario.inputType);
@@ -45,12 +91,11 @@ export async function buildVideoInput(
     }
   }
 
-  // For video generation, use clean prompt directly
-  // Modern models handle context natively
   if (typeof __DEV__ !== "undefined" && __DEV__) {
     console.log("[VideoStrategy] Using clean prompt for video generation", {
       promptLength: finalPrompt.length,
       photoCount: photos.length,
+      hasAudio: !!audioUrl,
     });
   }
 
@@ -61,6 +106,7 @@ export async function buildVideoInput(
     duration: extractDuration(wizardData),
     aspectRatio: extractAspectRatio(wizardData),
     resolution: extractResolution(wizardData),
+    audioUrl,
   };
 }
 

package/src/domains/generation/wizard/infrastructure/strategies/video-generation.types.ts

@@ -19,6 +19,8 @@ export interface WizardVideoInput {
   readonly aspectRatio?: string;
   /** Video resolution (e.g., "720p", "1080p") */
   readonly resolution?: string;
+  /** Audio file as base64 or URL for background music / audio-driven video */
+  readonly audioUrl?: string;
 }
 
 export interface CreateVideoStrategyOptions {
@@ -96,6 +98,10 @@ function isWizardVideoInput(input: unknown): input is WizardVideoInput {
     return false;
   }
 
+  if (obj.audioUrl !== undefined && typeof obj.audioUrl !== "string") {
+    return false;
+  }
+
   return true;
 }
 
@@ -137,6 +143,10 @@ export function validateWizardVideoInput(input: unknown): WizardVideoInput {
       errors.push("resolution (string, optional)");
     }
 
+    if (obj.audioUrl !== undefined && typeof obj.audioUrl !== "string") {
+      errors.push("audioUrl (string, optional)");
+    }
+
     throw new Error(`Invalid WizardVideoInput: ${errors.join(", ")}`);
   }
 

package/src/domains/generation/wizard/presentation/hooks/wizard-generation.types.ts

@@ -22,6 +22,8 @@ export interface WizardScenarioData {
   readonly description?: string;
   /** Video feature type - set by main app to control generation mode */
   readonly featureType?: "text-to-video" | "image-to-video";
+  /** Pre-generated image URL — when set, video generation uses this as source image instead of extracting from wizard data */
+  readonly preGeneratedImageUrl?: string;
   [key: string]: unknown;
 }